Skip to main content
Claude Code ships Dynamic Workflows: describe a task, Claude writes a JavaScript orchestration script, and a runtime fans it across parallel subagents attacking the problem from independent angles, converging on an answer. Max and Team plans already have it, genuinely good for work that fits in one sitting; you want Smithers when the work has to outlive it.

The short version

Dynamic Workflows is a feature of one vendor’s coding agent; Smithers is an MIT-licensed runtime treating coding agents, humans, and plain code as steps in one durable workflow. They overlap on “run many agents against a problem,” differing on everything else.

What that means in practice

Durability is the line. A Dynamic Workflow’s value lives only in that session; Smithers persists each completed step immediately, so killing the process, closing the laptop, or deploying over it costs nothing. Resume picks up at the right step; supervise revives stale runs automatically when a heartbeat goes stale. Vendor scope is the other line. Smithers outlives the model choice: a frontier model plans, a cheaper model fans out, and a different lab’s model reviews, all in one workflow, with fallback arrays (agent={[claude, codex]}) for a bad provider day or a rate limit. Dynamic Workflows orchestrates Claude only. Approvals change what you can automate. <Approval> suspends a Smithers run as a free database row and resumes when someone answers: the difference between agents you watch and agents you check on in the morning.

They compose

Smithers drives Claude Code as a first-class agent on your existing subscription; the Claude Code plugin mirrors its runs back into the session so your agent can watch and steer them. Split the work: quick fan-outs stay in the session, anything long-running, approval-gated, or worth rerunning goes to Smithers. More, with the crash-resume demo: the open, durable version of agent workflows.

See also